Neocortix was founded by Dr. Lloyd Watts to commercialize his 35-year research on Reverse-Engineering the Human Brain, which he began at Caltech in 1989, working with Silicon Valley pioneer Carver Mead.

In 2023, Dr. Watts recognized the Billion-Dollar Problem of Explainability and Hallucinations in Large Language Models (LLMs). Major technology companies were spending billions of dollars on LLMs, but were unable to solve the Explainability and Hallucination problem, thus jeopardizing their Trillion-Dollar valuations. Hallucinations were regarded by industry experts like Andrej Karpathy as an inherent property of LLMs, and LLMs and Deep Neural Networks (DNNs) were regarded as fundamentally unexplainable, so the problems were regarded as fundamentally unsolvable by all industry experts. Even Sergey Brin admitted in March 2024 that Google had no solution to the Hallucination problem and needed a breakthrough.

In early 2024, Dr. Watts had a series of breakthroughs that solved the Explainability problem for Deep Neural Networks and Large Language Models, by building a Deep Attribution Network to determine the best sample in the training dataset for the current context. This fundamental insight allows Large Language Models to explain why they are saying what they are saying, and to cite their sources from the training dataset. He also discovered a way to determine when Hallucinations are likely to occur, based on an internal confidence measure from the LLM Transformer model. This insight allows Hallucination Detection. Next, the Deep Attribution Network and Hallucination Detection methods can be combined and used to Mitigate Hallucinations, by detecting when Hallucinations are likely to occur and selecting tokens directly from the most relevant training document. And finally, the Deep Attribution Network technology can be used to determine Proportional Royalty Allocation, which is a solution to the problem of compensating creators whose work has been included in the training dataset. This is highly relevant to the New York Times vs. OpenAI copyright infringement lawsuit. Gary Marcus said at Stanford in September 2024 that there was no known technology that could do LLM Source Attribution and Royalty Allocation; that was before he had learned about Neocortix Deep Attribution Networks and LLM Source Attribution capability. Gary Marcus has subsequently joined the Neocortix Technology Advisory Board.

Dr. Watts has developed working code and real-time demonstrations for all 5 of the above inventions, and filed U.S. Patent Applications on all of them, and in late 2024 they are being evaluated at several of the world's largest technology companies.